On Thu, 31 May 2007 16:06:43 -0700, Edward Jaffe wrote:
>
>Mostly SAMxx. But, once again, I encourage you to run your own
>benchmark. Empirical evidence is the best kind.
Ok, I did. A small test program (below) with a million SAM64 and SAM31
instructions used 0.08 CPU seconds. When I changed the SAM* instructions
to LHI, it used 0.02 CPU seconds. This on a z890. I'd say that's pretty
expensive. Maybe not expensive enough to worry about though.
here's the source:
SAMTEST CSECT ,
SAMTEST AMODE 31
SAMTEST RMODE 31
SAVE (14,12)
USING SAMTEST,15
N 15,=X'7FFFFFFF'
LGFR 15,15
L 2,=F'1000000'
LOOP SAM64 ,
L 10,LR10
ST 10,STR10
C 2,=F'100'
BNE SAM31
NOP SAM31
SAM31 SAM31 ,
L 10,LR10
ST 10,STR10
BCT 2,LOOP
RETURN (14,12),RC=0
DS CL256
LR10 DC F'20'
STR10 DS F
LTORG
END SAMTEST
--
Tom Marchant
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
